Editorial Policy & Methodology

1. Read the document. Digital PDFs are read from their text layer. Scanned PDFs and photos are read with OCR. Password-protected PDFs are opened with the password you enter.
2. Identify the structure. An AI vision model locates the transaction table and the statement details: account holder, account number, statement period, and opening and closing balances.
3. Extract transactions. Each row becomes a transaction with date, description, debit, credit, and running balance, normalized to one date and number format.
4. Verify with the running balance. Starting from the opening balance, each transaction is applied and compared with the balance printed on the statement. Mismatches are flagged instead of silently corrected.
5. Export. The verified ledger is written to Excel, CSV, QBO, OFX, QIF, or PDF, using each format’s standard fields.

A bank statement is self-checking: every running balance must equal the previous balance plus credits minus debits. Recomputing that chain catches the most common extraction errors, such as a missed row, a debit read as a credit, or a misread digit, and shows you exactly where they are.

• Low-resolution, skewed, or heavily compressed scans reduce OCR accuracy.
• Handwritten notes and stamps over the transaction table may be misread.
• Statements without printed running balances cannot be fully verified with the balance check.
• Unusual layouts, merged cells, or multiple currencies in one table may need manual review.

Accuracy figures describe typical results on clean statements and are not a guarantee for any single document. We always recommend reviewing flagged rows. See the Disclaimer.

• Converter guides describe features that exist in the product today.
• Bank-specific pages describe statement layouts we support; bank names are used only for compatibility.
• Pages are reviewed when the product or a supported format changes, and dated when updated.
• We do not accept payment to feature or rank any bank or software.